Use results from one PostGIS SQL query in WHERE clause of another query

Question

I am building a mapping application that stores GPS tracks as lists of coordinates in a table called base_points.

I have a web application that uses leaflet to display these tracks. When I move the map, I make an AJAX call to my backend, with the bounds of the current view.

I'd like to return a simplified set of points within the bounds of the map.

At first, I used this query which works pretty well:

SELECT ST_AsGeoJSON(
               ST_Simplify(
                       ST_MakeLine(ARRAY(SELECT geom
                                         FROM base_point
                                         WHERE track_id = %s
                                           AND geom @
                                               ST_MakeEnvelope(
                                                       %s, %s, %s, %s, %s)
                                         ORDER BY time)),
                       %s
                   )
           );

The problem is that on certain zoom levels where the track exists and enters the viewport, I get a subset of the track as some parts are cut out by the geom @ ST_MakeEnvelop. I'd like to find the min and max track point within the viewport, and use those to retrieve all the points, and then using ST_Simplify(ST_MakeLine).

I am trying the following SQL expression, but it doesn't work.

WITH bounds as (SELECT MIN(id) as min, MAX(id) as max
                FROM base_point
                WHERE track_id = %s
                  AND geom @ ST_MakeEnvelope(%s, %s, %s, %s, %s))
SELECT ST_AsGeoJSON(ST_Simplify(ST_MakeLine(ARRAY(
        SELECT geom 
        FROM base_point, bounds 
        WHERE track_id = %s 
            AND id < bounds.max 
          AND id > bounds.min ORDER BY time
    )),
%s)); 

I've edited this from earlier, and the query now works, but is very slow.

Here's an explain/analyze output:

Result  (cost=8.95..34.46 rows=1 width=32) (actual time=59806.894..59806.896 rows=1 loops=1)
  InitPlan 1 (returns $2)
    ->  Sort  (cost=8.95..8.95 rows=1 width=40) (actual time=59803.755..59804.242 rows=14001 loops=1)
"          Sort Key: base_point.""time"""
          Sort Method: quicksort  Memory: 1478kB
          ->  Nested Loop  (cost=4.88..8.94 rows=1 width=40) (actual time=13.074..59799.753 rows=14001 loops=1)
                Join Filter: ((base_point.id < (max(base_point_1.id))) AND (base_point.id > (min(base_point_1.id))))
                Rows Removed by Join Filter: 523
                ->  Index Scan using base_point_track_id_b527d95c on base_point  (cost=0.43..4.45 rows=1 width=44) (actual time=0.031..2.440 rows=14524 loops=1)
                      Index Cond: (track_id = 40)
                ->  Aggregate  (cost=4.46..4.46 rows=1 width=8) (actual time=4.117..4.117 rows=1 loops=14524)
                      ->  Index Scan using base_point_track_id_b527d95c on base_point base_point_1  (cost=0.43..4.45 rows=1 width=4) (actual time=0.005..3.477 rows=14003 loops=14524)
                            Index Cond: (track_id = 40)
                            Filter: (geom @ '0103000020E61000000100000005000000C3F5285C8FF256C08FC2F5285C4F4340C3F5285C8FF256C07B14AE47E15A4440A4703D0AD73355C07B14AE47E15A4440A4703D0AD73355C08FC2F5285C4F4340C3F5285C8FF256C08FC2F5285C4F4340'::geometry)
                            Rows Removed by Filter: 521
Planning Time: 0.294 ms
Execution Time: 59807.087 ms

If I separate out the queries and hard code the ids for comparison, the query runs in 15ms.

Answer 1

I figured it out. Doing a join is much faster.

WITH bounds as (SELECT MIN(id) as min, MAX(id) as max
                FROM base_point
                WHERE track_id = %s
                  AND geom @ ST_MakeEnvelope(%s, %s, %s, %s, %s))
SELECT ST_AsGeoJSON(ST_Simplify(ST_MakeLine(ARRAY(
        SELECT geom 
        FROM base_point
        INNER JOIN bounds ON
            id < bounds.max 
          AND id > bounds.min ORDER BY time
    )),
%s)); 

If someone can explain why, I'd greatly appreciate it!